Openai today sent another earthquake rocking in the world of software engineering.
This time, it was X declaration This Codex, its AI coding agent, is now available to plus tier users. In other words, now you will not have to spend $ 200/Mo to get Codex’s programming assistance. Instead, you can get it for $ 20/Mo, as part of the chatup plus membership.
(Disclosure: ZDNET’s original company Ziff Davis filed a case of April 2025 against Openai, alleging that it violates Ziff Davis copyright training and operating its AI system.)
The company says that Codex “includes limited time limit boundaries, but during the period of high demand, we can set a rate limit for plus users so that the Codex is widely available.” In other words, today performance may be thicker.
Also: Codex with Openai upgrade codgpt – and I am seriously impressed (so far)
This did not stop me from digging. The Chatgpt is a new menu item on the right side of the screen:
Interestingly, the MAC app version of Chatgpt is far more limited. If you launch Chatgpt through the app, not only will you get Codex, but also the search chat, library and Sora are missing. This is an unexpected error, especially since the Mac app was updated yesterday.
Also: Best AI for coding in 2025 (and what not to use)
In any case, let’s see what happens when you click on the Codex button in Plus. First of all, you will get this exciting start screen.
Next is an animation that describes some characteristics of the device. In this first frame, the equipment suggests that it will draft several PRS in parallel. The PR stands for the push request, and it is github-express to draw changes in the main repository, essentially integrates it into the main codebase.
Codex draft by saying several PRS, it means that it can work on many coding projects, but no one is committed until you have reviewed. This is roughly equivalent to giving a group of projects to some subordinate programmers, allowing them to work on it, but do not include their code until you check it carefully.
You will carefully examine the code of Codex, will you not? prompt. you better.
Next is a sample signal. This is basically showing how you can motivate the codex to dig the degrees through the entire repo of the codebase and to tweak.
Next, Codex can clearly run lint and test. Lint, which was originally a C language tool that detects errors and bugs in the code, works with many different languages.
What Codex is saying here is that this coding verification is capable of running checks, making tests and running tests that are required to run tests to confirm the work of code changes. This is a good thing.
Finally, another panel describes the use of a new coding model specifically for software engineering. The company clearly does not specify the model, but Chatgpt estimated that the model is a tuned version of the GPT-4o, which is particularly trained on software engineering applications.
And then, I was finally inside. The next form provided a connection to a githib repository. This reminds me of the process that I went with Google Jules coding model last week. I am telling you, these AI features are in lockstap among these contestants.
I was about to move forward and started playing with Codex when I saw a message under the screen.
It refers to a feature mentioned in X posting by Openai this morning:
Now you can give codax access to the Internet during task execution to establish base dependence, run tests that require external resources, upgrade or install the package required for the construction of new features, and more.
Stay for testing on hands
And when I decided that it was time to stop. It is not even at 7 am here, I have not even finished my first cup coffee, and I want to get this news out as soon as possible. But I am not loosening Codex on comfortable, foggy morning head and all, on my main repo, or thinking through the implications of giving it an internet access.
Also: Google’s Jules AI coding agent created a new feature that I could actually ship – while I made coffee
He said, I will share an interesting search with you, while Caffeine was starting to promote my brain. Obviously, Codex is the stateless. This means that it does not remember anything from session to session. Therefore you need to create a clear and full signal, and make sure those signals are repeated in new sessions if they establish land rules or otherwise are designed to check Codex at its work.
And with this, I will wrap this quick hit intro article. I will soon come back with some hands tests as to what this animal can do. Stay
Are you planning to try the new Codex coding agent in Chatgpt Plus? What are your views on letting the Internet join the Internet to establish dependence? Do you feel comfortable to provide it to your codebase, or are you worried about automation at this level? What features are you the most excited or nervous about? Let us know in the comments below.
Want more stories about AI? Sign up for innovationOur weekly newspapers.
You can follow my day-to-day project updates on social media. Be sure to subscribe to My weekly update newsletterAnd follow me on Twitter/X @DavidgewirtzOn Facebook Facebook.com/davidgewirtzOn Instagram Instagram.com/davidgewirtzOn blue @Davidgewirtz.comAnd on youtube Youtube.com/davidgewirtztv,
